The paper reconsiders the theory of fiscal federalism in a framework inspired by property-rights theory. We set up a two-period model where on a first stage a region in a federation can expend value-enhancing investments into a public project. The project can be implemented on a second stage, and causes spilovers on other regions. Under centralized as well as decentralized governance, negotiations on the federal level facilitate the realization of the efficient policy. Still, non-contractibility of investments causes the overall outcomes to differ across regimes. If the region with access to the public project bears the entire implementation costs of its policies, underinvestment prevails and subsidiarity (centralized governance) is superior when spillovers are weak (strong). Conversely, if linear cost-sharing arrangements are feasible, decentralized authority often leads to a socially optimal outcome while centralized authority (with majority or unanimity rule) does not.